Appearance
Smart Assist: plans, allowances and limits
Last updated
AI features are a Scada.Ink Cloud service. They are not unlocked by the appliance's license tier; the cloud decides on each request.
Checks on every request, in order
- License key — present and valid, signed by Scada.Ink, locked to an appliance's hardware ID, not expired, not revoked.
- Rate limit — requests per minute.
- Plan — the appliance is registered to a Scada.Ink Cloud account with an active AI plan (active, trialing or payment overdue all count). Deep analysis needs a Pro plan.
- Request size — within the limits below.
- Allowance — the account has requests left this month.
Plans that include AI
| Plan | Billed | AI analysis requests / month | AI page designs / month | Deep observation analysis |
|---|---|---|---|---|
| Smart Assist | Per account | 300 | 20 | No |
| Smart Assist Pro | Per account | 1,500 | 100 | Yes |
| Cloud Complete | Per appliance | As Smart Assist | As Smart Assist | No |
| Cloud Complete Pro | Per appliance | As Smart Assist Pro | As Smart Assist Pro | Yes |
Prices and subscriptions are managed in the Scada.Ink portal. If an account holds more than one AI plan, the highest limits apply; they are not added together.
How the allowance counts
- One request is one unit, whatever its size — 5 tags or 200, a 3-stage deep analysis or a quick check.
- There are two pools, per Cloud account (shared by all its appliances): Smart Assist analysis requests (AI-enhanced classification and deep analysis) and AI page designs.
- In the page designer, every round is one design: the first Design the page, each Continue after questions, and each Apply of a change.
- A classification or deep analysis that fails is not counted. A page design that reached the model is counted even if the answer could not be used.
- The allowance runs for the calendar month in UTC and resets at 00:00 UTC on the 1st.
- When it is used up: "Monthly allowance used: this account has used all 300 Smart Assist analysis requests included in its Scada.Ink Cloud plan for September. The allowance resets on 1 October (UTC); upgrade the plan in the Scada.Ink portal for more."
Cost estimates on the page
The estimated $ figures on the Enhance classification with AI? and Deep AI analysis cards are rough estimates of the AI model's processing cost, worked out in your browser from the number of tags. They are not your bill: under a plan, each request uses one allowance unit. They use the list prices of the models the cloud uses (Claude Haiku 4.5 for classification, triage and verification; Claude Sonnet 5 for the deep classification stage).
Rate limits
| Requests | Limit |
|---|---|
| Classification and deep analysis | 60 per minute per appliance |
| Page design | 6 per minute per appliance, and 30 per minute across all appliances |
Over the limit: "Rate limit exceeded. Maximum 60 requests per minute." or "This service is busy. Try again in a minute." (429).
Request size limits
| Request | Limit |
|---|---|
| AI-enhanced classification | 200 tags |
| Deep analysis | 1,000 tag fingerprints (the appliance collects at most 500) |
| Page designer | 600 tags across all chosen devices (shared evenly, smallest device first); 20 devices; description 4,000 characters; page being edited 200,000 characters; 10 answers of up to 1,000 characters; tag fields 200 characters (longer tag paths are left out) |
| Any request body | About 1 MB |
Not included
- The AI features do not work without a registered appliance and an AI plan, even on a trial license with no Cloud account.
- There is no usage page in the appliance. The cloud keeps per-appliance totals; check usage in the Scada.Ink portal.